RAP Reveals Buccaneers' Secret Weapons

Tampa Bay Buccaneers C Ryan Jensen
Tampa Bay Buccaneers C Ryan Jensen
Photo: USA Today Sports Images

Guest column by Ajit Kirpekar

The Super Bowl Champions had a lot of stars on offense. Tom Brady set a team record with 40 touchdown passes. Mike Evans caught 13 of those scores, also a franchise-best. Leonard Fournette racked up nearly 450 yards from scrimmage in the playoffs alone. The best players on that Buccaneers offense, however, may have been the linemen: Donovan Smith, Ali Marpet, Ryan Jensen, Alex Capp, and Tristan Wirfs.

Welcome to the 2020 edition of Rate of Adjusted Pressure, or RAP for short. For those unfamiliar, RAP is a statistic that measures a team's ability to generate or avoid pass pressure, opponent adjustments included.

As a refresher, RAP is calculated by taking a set of inputs that correlate well with pressure—e.g., down, distance, field location, etc.,—and running them through a machine learning model to predict the probability of pressure. From there, the model is rerun with a subset of opponent-related inputs set to the average, thus giving us an opponent-adjusted rate of pressure. For a more detailed explanation, see this link. And for last year's article, click here. Reminder, all data comes from Football Outsiders' game charting data (collected first by FO staff and volunteers, then in conjunction with ESPN Stats & Info, and since 2015 by Sports Info Solutions).

As is typical, every offseason the stat goes through another year of iteration. It's important to note that with each iteration, the adjusted numbers shift slightly. This year, in addition to some machine learning-related improvements, the big change includes quarterback time to throw as an input into the model. As one can imagine, a quarterback's time to throw is going to heavily correlate with the probability of a team generating or giving up pressure, making it a useful addition. Unfortunately, time to throw is a relatively recent statistic and is only publicly available through NFL Next Gen Stats since the 2016 season. In contrast, FO's game charting goes back all the way to 2006.

In order to get around this issue, I created an expected time to throw statistic that works in the following way:

  1. Estimate the time to throw from 2016 to 2020 using data available during those seasons. The variables in the model include shotguns called per game, play-actions, depth of passes, etc.
  2. Run the model against data from 2006 to 2020, providing an estimate for every quarterback in the sample.
  3. Use the estimated time to throw in the RAP model.

With that out of the way, what I like to do next is take a closer look at some big offseason changes that occurred from last year. Usually these involve trades, such as Houston's deal for Laremy Tunsil or Chicago's acquisition of Khalil Mack. This year, however, we are going to take a look at the New England Patriots, who for the first time in eons had a new starting quarterback under center. Even more intriguing, the incoming quarterback was about as different stylistically from the outgoing quarterback as you could imagine. So how did New England perform?

 

New England Patriots RAP, 2019-2020
Season RAP Pressure
Rate
Rank
2019 20.2% 25.0% 3
2020 20.0% 22.2% 5

As it turns out, even with a completely different quarterback under center, New England essentially had the same RAP. The results were so puzzling that I took a deeper look at both quarterbacks.

Cam Newton Pressure Rate Year-by-Year
Year Pressure
Rate
Time To
Throw
Estimated
Time To Throw
2011 24.0% - 2.64s
2012 26.0% - 2.69s
2013 25.7% - 2.64s
2014 25.2% - 2.68s
2015 27.9% - 2.70s
2016 29.6% 2.69s 2.66s
2017 29.6% 2.73s 2.67s
2018 27.7% 2.67s 2.66s
2019* - - -
2020 21.3% 2.75s 2.69s
* Missed 14 games due to injury

Prior to last year, Cam Newton's pressure rate and time to throw statistics had been relatively consistent. And yet, much like Teddy Bridgewater a year ago, Cam Newton on a different team replacing another Hall of Fame legend had the best season of his career, at least by this metric.

For comparison, check out Tom Brady's historical results.

Tom Brady Pressure Rate Year-by-Year
Year Pressure
Rate
Time To
Throw
Estimated
Time To Throw
2006 16.1% - 2.44s
2007 14.9% - 2.37s
2008* - - -
2009 18.9% - 2.43s
2010 19.0% - 2.47s
2011 22.2% - 2.50s
2012 16.5% - 2.50s
2013 20.7% - 2.50s
2014 22.7% - 2.53s
2015 24.0% - 2.52s
2016 26.7% 2.59s 2.52s
2017 28.4% 2.71s 2.61s
2018 25.9% 2.62s 2.60s
2019 25.0% 2.73s 2.68s
2020 15.0% 2.57s 2.61s
* Missed 15 games due to injury

As you can see, Tom Brady himself had his own tale of two seasons, which we'll get into in the next section when we dive into the best offenses.

2020 Offenses

Lowest Offensive RAP, 2020
Rank Offense RAP Pressure
Rate
Difference
1 PIT 14.7% 14.9% +0.2%
2 TB 14.9% 15.0% +0.1%
3 NO 15.1% 17.7% +2.6%
4 IND 17.2% 18.4% +1.2%
5 NE 20.0% 22.2% +2.2%
6 LV 20.7% 21.1% +0.4%
7 GB 20.8% 22.2% +1.4%
8 TEN 21.4% 21.3% -0.1%
9 WAS 21.7% 23.5% +1.8%
10 BAL 21.8% 26.4% +4.6%
12 DET 22.1% 21.2% -0.9%
11 CIN 22.1% 22.2% +0.1%
13 ARI 23.0% 23.0% +0.0%
14 CLE 23.2% 24.6% +1.4%
15 MIA 23.5% 23.7% +0.2%
16 CHI 23.6% 25.8% +2.2%
17 LAR 23.7% 24.6% +0.9%
18 LAC 24.0% 26.5% +2.5%
19 JAX 24.1% 25.2% +1.1%
20 DAL 24.3% 24.2% -0.1%
21 CAR 24.8% 24.0% -0.8%
22 ATL 26.0% 26.1% +0.1%
23 BUF 26.3% 29.6% +3.3%
24 KC 27.2% 28.5% +1.3%
25 PHI 27.3% 31.6% +4.3%
26 DEN 28.7% 29.5% +0.8%
27 SF 29.4% 27.9% -1.5%
28 SEA 29.5% 30.9% +1.4%
29 HOU 30.0% 29.9% -0.1%
30 NYG 30.2% 29.2% -1.0%
31 NYJ 30.7% 32.6% +1.9%
32 MIN 31.8% 31.6% -0.2%

The 2020 season was a career renaissance for Brady. Not only did he extend his monopoly rights over the Lombardi Trophy, he enjoyed his best season of RAP in over a decade. After a year when scores of people, including yours truly, proclaimed Brady ready for the coffin, he once again proved to his doubters that he's an unkillable vampire.

On a more serious note, Brady's stunning reversal in RAP, along with Newton's own outlier season, suggests context around the quarterback plays a significant role in a team's RAP.

As for the Steelers, this was Ben Roethlisberger's best season in terms of RAP, in a career that has been wildly inconsistent.

Ben Roethlisberger Pressure Rate Year-by-Year
Year Pressure
Rate
Time To
Throw
Estimated
Time To Throw
2006 20.7% - 2.60s
2007 35.7% - 2.89s
2008 24.2% - 2.63s
2009 23.6% - 2.55s
2010 34.3% - 2.58s
2011 30.2% - 2.61s
2012 24.3% - 2.63s
2013 24.4% - 2.55s
2014 18.4% - 2.51s
2015 14.4% - 2.49s
2016 19.2% 2.60s 2.50s
2017 23.6% 2.59s 2.51s
2018 24.2% 2.53s 2.47s
2019* - - -
2020 14.2% 2.38s 2.51s
* Missed 14 games due to injury

Near the bottom of the table, we find the New York Jets, who finished dead last a year ago. As the old adage says, "when it gets really bad, it's hard to get much worse," which proved to be true in the literal sense as they catapulted all the way from 32nd to 31st. "A day late and a dollar short."

The other New York team didn't fare much better, although I suppose it's some consolation that they managed to one up their crosstown rivals yet again. This was yet another dismal showing; two straight years spent living in the toilet.

Minnesota's last-place finish should come off as a major surprise, even though it may not feel that way. In their ugly 2019 playoff loss, their pass protection got completely dominated by the 49ers. However, in 2019 they were actually a (very) slightly above average team in rank. As such, their 2020 pratfall is a rather stunning result.

A final word on the offenses before we move onto the defenses. Every year under Russell Wilson, the Seahawks have finished in the bottom 10 in RAP. And every year, the Seahawks manage to once again find themselves among the very worst teams in this statistic. While hope springs eternal, it's hard not to imagine peeking at this list a year from now and seeing Seattle listed at the bottom. If it's any consolation, it appears the Seahawks will not be alone. Just as with Wilson, every year under Deshaun Watson the Texans have also finished at the bottom of these rankings. We went over the Tunsil trade a year ago, and a year later the results continue to be grim.

2020 Defenses

Highest Defensive RAP, 2020
Rank Defense RAP Pressure
Rate
Difference
1 TB 33.3% 28.8% +4.5%
2 LAR 30.1% 30.5% -0.4%
3 NO 29.6% 29.7% -0.1%
4 KC 29.1% 29.0% +0.1%
5 PIT 28.4% 29.4% -1.0%
6 SEA 28.1% 23.8% +4.3%
8 WAS 26.7% 27.1% -0.4%
7 MIA 26.7% 28.3% -1.6%
9 TEN 26.2% 22.6% +3.6%
10 NE 25.6% 28.9% -3.3%
11 BUF 25.1% 25.2% -0.1%
12 ARI 25.0% 25.2% -0.2%
13 BAL 24.6% 26.2% -1.6%
14 SF 24.3% 26.1% -1.8%
15 PHI 23.8% 28.8% -5.0%
16 NYG 23.0% 24.8% -1.8%
17 CLE 22.7% 23.6% -0.9%
18 IND 22.5% 24.1% -1.6%
19 CAR 22.4% 24.3% -1.9%
20 ATL 21.5% 24.2% -2.7%
21 CHI 20.7% 22.8% -2.1%
22 LAC 20.1% 22.4% -2.3%
23 DEN 19.8% 24.0% -4.2%
24 LV 19.7% 23.0% -3.3%
25 HOU 19.3% 20.0% -0.7%
26 NYJ 19.3% 24.3% -5.0%
27 GB 19.1% 23.1% -4.0%
28 DAL 19.0% 26.6% -7.6%
29 CIN 18.3% 20.0% -1.7%
30 JAX 17.9% 19.8% -1.9%
31 MIN 17.0% 18.7% -1.7%
32 DET 16.9% 20.5% -3.6%

Here's what I wrote about Miami in 2019:

Did I mention it was a bad year for Miami? No? Well, it was bad. It was really bad. We don't have to talk about that anymore; it was just bad. That's it. Hopefully next year will be better.

I don't know if even the most ardent Dolphins fan could have predicted this level of a turnaround. It's not that it was an impossibility (obviously), but it's nonetheless on the extreme range of outcomes. And yet they have company in the Seahawks, who went from an abysmal 30th in RAP a year ago to sixth, though a big chunk of that is owed to opponent adjustments.

Interestingly, quite a few teams repeated their performances from a year ago, including the Rams, Steelers, Saints, and Patriots. For the Patriots, the ranking is unusual, especially considering that a year ago the defense was regarded as one of the all-time great units. A year later, when the defense was ravaged by defections, COVID, and opponent adjustments … they still managed to finish in the top 10.

Also deserving of major plaudits here are the Rams. For a team who is known for their offense, their defense has managed four straight seasons in the top five, including two first-place finishes and a second-place finish. No other franchise in the NFL over the last 15 years has accomplished this.

If there's a photo negative of the Miami Dolphins, it's the Jacksonville Jaguars, who took their plucky, feel-good season from a year ago and landed right into the gutter. It's hard to overstate just how far they fell in one year—the largest single-season collapse in the entire sample.

Minnesota's defensive collapse mirrors their offense almost exactly, finishing (very) slightly above average one year ago and then nose-diving a year later; although the turnover in defensive personnel helps explain this result.

And lastly, the Detroit Lions did their usual Detroit Lions thing, falling from embarrassing a year ago to abominable. Maybe a Miami-like turnaround is in the cards next year?

Ajit Kirpekar is a data scientist based out of San Francisco, California. Despite this, he roots for the Indianapolis Colts.

Comments

4 comments, Last at 10 Aug 2021, 8:23pm

1 Peyton Manning

Peyton Manning had an extra 1.5 seconds in the pocket when Jeff Saturday was protecting him. Without Saturday, when Peyton played for the Colts, Peyton’s stats went way down due to his immobility and being pressured. Totally different QB without Saturday.

2 Unfortunately, time to throw…

In reply to by Rekroot

Unfortunately, time to throw was not available for any of Manning's career. But I can give you estimated time to throw for him

 

2006 - 2.49

2007 - 2.48

2008 - 2.43

2009 - 2.39

2010 - 2.36

2012 - 2.36

2013 - 2.38

2014 - 2.40

2015 - 2.47

Here are his pressure rates(raw)

2006 - .11

2007 - .14

2008 -.14

2009 -.15

2010 - .14

2012 - .14

2013 - .15

2014 - .13

2015 - .21

3 A word on estimated time to…

A word on estimated time to throw. The model fit stats were just OK. By that I mean, the range the estimates are "off" is typically withing half of a second in either direction. And I tried this model through multiple different algorithms. 

Overall I was happy with it for two reasons

1) There was no bias in the estimates. Meaning, if it was off, it wasn't going to be systematically off for one type of QB vs another

2) The results preserved the ordinal ranking of the QBs. Meaning, the passers who finished with low time to throws were the same in the estimated model along with those that finished high in the time to throw stat.